Uncoupled and coupled hydrogeophysical inversions of seawater intrusion and geothermal hydrologic models

Most of humans live in coastal and alluvial plains. Sustainable water supply systems are
essential for drinking water production while long-term groundwater access ensures its geothermal
use. Owing to climate change and anthropogenic activities, groundwater resources will be
increasingly threatened throughout the twenty-first century. In the context of over-development,
any effort to overcome groundwater mining at the expense of better understanding groundwater
management problems may be misguided. A combination of robust measuring technologies and
reliable predictions based on numerical models are necessary to estimate better hydrogeological
parameters. Sparse and continuous data are increasingly being used conjunctively in hydrogeological
modeling and inverse calibration to alleviate extrapolation and subjective interpretation.
In hydrogeology, electrical resistivity tomography (ERT) gives quantitative and qualitative information
on salinity and temperature distribution. ERT offers a means for mapping the subsurface
hydrogeological structures and the hydrologic process dynamics in both space and time.
The integration of ERT data to hydrogeological calibration is particularly suited to define the
hydrogeological models quantitatively and qualitatively. In a deterministic framework, hydrogeophysical inversion methods may be particularly useful for predictive analysis and assessment of groundwater pre-emptive management strategies.
This work focuses on calibrating physically-based and spatially distributed groundwater flow
and transport models with non-to minimally invasive electrical tomography. The general objective
is to provide better estimates of hydrogeological parameters on two relevant problems:
seawater intrusion in coastal areas and geothermal energy in shallow environments. We compare
methods to integrate quantitatively ERT data, possibly other hydrologic or geological data,
with existing information to hydrogeological models. Two hydrogeophysical inversion schemes
are developped and a thorough comparison of an uncoupled and a coupled quantitative approach
based on the use of surface ERT only is performed. The uncoupled hydrogeophysical inversion
involves constraining hydraulic parameters using geophysically-derived data and first requires a
geophysical inversion, after which the geophysical parameters are converted to hydrologic data
through a petrophysical relationship. The inverse hydrological calibration is then performed
on these inferred hydrological data. The coupled hydrogeophysical inversion involves constraining
hydraulic parameters using geophysical data through a forward hydrogeophysical model in
which the hydrologic data are converted to resistivities through a petrophysical relationship. A
geophysical forward problem is then solved for the geophysical data. The inverse hydrological
calibration is performed on the inferred geophysical observations. In both schemes, we show that an independent geophysical inversion is required to delineate heterogeneous bodies. In
this context, we study how to derive informative content of ERT images and therefore ERTderived
hydrologic data and ERT-derived geometry. We show that a quantitative appraisal (the
cumulative sensitivity) must be used as a proxy for filtering areas correctly resolved.
Our developments are demonstrated on several benchmark SWI models (numerical and analytical),
a thermohydrologic model and a field test. We show that the reliability of estimated
SWI model parameters with the uncoupled approach depends on ERT image appraisal, on
geophysical data collection strategy and hydrogeological model conceptualization. The ERT
image appraisal plays a key role in retrieving high quality ERT-derived data and helps discriminate
different measurement arrays. It is particularly useful in preventing the integration
of noise-related artefacts in the conceptualization. We endeavor to quantify the modeling error
by a thorough comparison of different strategies to assess the effect of decreasing model
conceptual errors to hydrogeological calibration. In the SWI analytical models, we highlight
the subjectivity of the uncoupled approach due to the nature of the required hydrologic data
(sharp interface). We demonstrate that the conjunctive use of an image appraisal tool with the
well-known Ghyben-Herzberg solution is needed to infer reliable ERT-derived observation data.
We further demonstrate that a SWI analytical solution may be used to calibrate an equivalent
hydraulic parameter based on an ERT dataset generated from a density-dependent groundwater
flow model. In the thermohydrologic model, we show that the effectiveness of the uncoupled
scheme in calibrating heat transport parameters may be hampered due to the regularization
constraint in the geophysical inversion. We demonstrate the importance of a noise level-related
filter on the time-lapse ERT images aimed at properly quantifying the spatio-temporal ERTderived
temperature changes. We also advocate the use of a physically-based constraint on the
ERT-derived temperatures to account for spatial mixing of waters and to cope indirectly with
the smoothing effect in the ERT images. In each application, the coupled approach significantly
prevails over the uncoupled scheme in terms of reliability of the parameter estimates when no
model conceptual error exists.